I've revisited some comments made concerning how artificially robotic acapella multitracks sound when I use Melodyne to create vocal harmonies.   I agree with that assertion.  Moreover I was told that if I want to persist and use Melodyne  to create harmonies then before so doing, I should spend the time learning the ins and out of Melodyne. I still have much to learn, but while having attempted to follow that sage advise vocal harmonics remain artificial sounding. I may have been told that Melodyne will never be able to replicate live vocal.  However, the demonstration that Celemony uses to describe the feature that creates vocal harmonies leads one to believe that it is a fairly simple matter.  What am I missing?   
 
My question to forum members  is this!  Can Melodyne satisfactorily replicate live vocals ... in particular live vocal harmonies  over an entire composition?  Does use of of EQ and Compression etc (about which I have little competence) help a lot in creating realistic Melodyne vocals!
 
 

Two things. However good Melodine is, you WILL get artefacts. Humans are so used to hearing real voices, the slightest artefact breaks the magic.

Harmonies created from a melody line will match too perfectly. Again, it sounds false simply because it is too perfect.


The trick to faking harmonies is to not do it... What I mean is, have a loose go at singing close to what you want, then use Melodine to tune it to the harmony you intended. You gain all the timing errors that differentiates between the voices, and you're minimising the tuning artefacts to help hide the fact that it is false.

All the harmonies in these songs were generated by Melodyne...judge for yourself if they’re convincing. If you think they are, I’ll tell you what I did to make them so. If not, I really don’t have anything else I can offer...I think they’re pretty convincing, though.
 
This one uses Melodyne to create massed vocals in the very beginning (using Take Lanes, in case you wondered). Then there aren’t any harmonies for a while, they’re used for embellishments at the end of lines in the verses.
https://www.youtube.com/watch?v=F33jgLw-uJI
 
This probably exposes Melodyne the most in terms of having to stand by itself. The main harmonies start kicking in around 1:00.
https://www.youtube.com/watch?v=lvLNkdGv4Kg
 
This is loaded with Melodyne harmonies. There’s only one “real” vocal I actually recorded in it.
https://www.youtube.com/watch?v=RuE55rwiQ9s
 
This one has a bunch of harmonies kicking in at 2:00. The “female backup harmonies” during the guitar solo used Melodyne’s formant function to get a higher-pitched character, and all the background vocals are Melodyned.
https://www.youtube.com/watch?v=fsp5hOxRX1M
 
In the video I did at Berklee last March, there’s a section where I added a vocal harmony to a cover version of “Can’t Explain.” Skip ahead to 5:30 for the section on Melodyne harmonies.
https://www.youtube.com/watch?v=ySzza5AWe0o
 
That should give you an idea of what's possible...

Over processed digitally altered sound of any kind become unnatural to human listeners because it's not what we are used to hearing in the real world. The sound will always be wrong because of the way our brains work.
Hit a bell and everyone will envision a bell. Record the bell and change the pitch a little bit and most will still say  that's a bell. Change the pitch to much and people will go WTF is that?
 So you need to fool peoples ears into thinking they are hearing what your intending them to hear by applying just the right amount and never too much. If you push the limits then other acoustical treatments like reverb might mask it, but even that will fail eventually. 
Here in lies the danger of pitching tools like melodyne. They will work fine for small amounts of pitch change but things become strange beyond a certain point. 
I find you can use digital harmony if it's back in the mix and a little more reverb than I would like. For in your face dry vocals nobody and no software can fool the human ear.  
 
So what I sometimes do is try my best to sing that higher part as close as I can and then turd polish it with pitch correction. If I make a pitch mistake on a lead vocal, I re sing it,,, faster and much better results. I only use pitch correction on harmony and the odd client who keeps hittin the same wrong note. 

+1 to the above. Melodyne works best when given clean (no effects whatsoever) audio data, and asked to process that data by small amounts. That said, it is much more effective to (sparingly) correct random misses on a harmony that was sung, rather than shifting a melody 4-5 semitones to "create" a harmony from a melody line.

  

Karyn
The trick to faking harmonies is to not do it... What I mean is, have a loose go at singing close to what you want, then use Melodine to tune it to the harmony you intended. You gain all the timing errors that differentiates between the voices, and you're minimising the tuning artefacts to help hide the fact that it is false.
 

Anderton
All the harmonies in these songs were generated by Melodyne...judge for yourself if they’re convincing. If you think they are, I’ll tell you what I did to make them so. If not, I really don’t have anything else I can offer...I think they’re pretty convincing, though.
 

Thanks you Karyn and Craig.  
 
 
Karyn
I have in fact "faked" harmonies as you suggest, using Melodyne simply to tweak the pitch structure into the desired result. Sometimes I even record a harmony two or three half-steps lower (to get within my vocal range) make Melodyne adjustments,  then transpose back up to the proper key after which I adjust formant.  Seems to work OK.  
Craig,
I find the harmonies convincing as they exist with the rest of the voicing an instrumentation.  I really hear little artificiality.  Then again my 73 year old ears may not offer me the best judgement. I am curious how the harmonies sound when listened to as solo.  Of course that helps bring out the "robotic" nature  if is exists at all. I again watched the Berklee demo (gosh was that way back in March?) I think I am doing  things pretty much the way you demonstrated in that segment of the tutorial.
 
Please don't burden yourself but I would certainly appreciate any advise you are willing to share that makes the harmonies sound more convincing.  I sing in the 2nd bass to baritone range and sometimes find myself wanting to create Tenor Alto and Soprano  harmonies. Consequently  it may be that my chief issue lies in my trying  (using Melodyne) to move pitch as too  far up the scale ... an octave or maybe more.  
 
I've found some success in using Process:Transpose   to move the tune,  for example, to the nest higher octave. From there it at least seems somewhat more convincing to dial-in a female sounding vocal by using the formant tool. Does that make sense?    

cpkoch
Please don't burden yourself but I would certainly appreciate any advise you are willing to share that makes the harmonies sound more convincing.  I sing in the 2nd bass to baritone range and sometimes find myself wanting to create Tenor Alto and Soprano  harmonies. Consequently  it may be that my chief issue lies in my trying  (using Melodyne) to move pitch as too  far up the scale ... an octave or maybe more. 

 
More than a fifth is really pushing it, as far as I'm concerned. 
 
But for smaller intervals, don't overlook the importance of using the formant tool, which is unfortunately not in the Essential version. If you transpose something up three semitones, using formant to bring the timbre back down a bit makes all the difference in the world. It was the formant tool that made the background vocals in "Maladie Du Coeur" sound like it was other people singing with me.
 
A sneaky but effective trick is to slide the synthesized harmony a little later than the real vocal. The ear is most interested in the attack of a sound ("Is that the wind, or a saber-tooth tiger?"). If the real vocal monopolizes the listener's attention and the harmony comes in with more ambience, timing differences, and a lower level, it will be gone before the next chunk of "real" vocal appears.
 
Melodyne's timing tool is great for shifting the initial part of generated harmonies behind the main vocal. 
 
You also have to be careful about how much vibrato you remove, although again this is an Editor version tool...not sure which version you have.

All female harmonies are Melodyne generated.  Christian contemporary....as a *heads up*  These vocals are fairly exposed as well...iow you can really hear them...they're not heavily buried in the mix. I purposely didn't mess much with the format tool, as I was liking the difference from the main vocal.  Pretty much a learning curve for me on this tune as well.
https://www.youtube.com/watch?v=6gJXOAEVFPA

Sock .. Mike
 
Thanks to you both.  I think I got the message and it is not unexpected.    Just curious though ... what do you think of using Process:Transpose  to raise the key and then tweak the pitch and formant using Melodyne? I  assume Melodyne's  Pitch change and Sonar's Transpose use different algorithms.  When combined they might well produce better results!  I think that is what I have found although admittedly with limited experimentation.

Oh, and one more thing...humans don't gravitate to even-tempered pitch but to just intonation. If you don't believe me, run an analysis on string ensembles and barbershop quartets, neither of which is constrained to even-tempered tuning. Often after creating a quantized harmony, I'll open it, turn off pitch grid, and make some slight adjustments to create what "sounds" right.